Skip to main content

Scaffolding for the Global Function Search optimizer from Dlib

Project description

gfsopt

Documentation Status Latest Version

pip3 install --user gfsopt

Convenient scaffolding for the excellent Global Function Search hyperparameter optimizer from the Dlib library. (See: 'A Global Optimization Algorithm Worth Using')

Provides the following features:

  • Parallel optimization: Run multiple hyperparameter searches in parallel
  • Save and restore progress: Save/restore settings and optimization progress to/from file.
  • Average over multiple runs: Run a stochastic objective function using the same parameters multiple times and report the average to Dlib's Global Function Search. Useful in highly stochastic domains to avoid biasing the search towards lucky runs.

Example usage

A basic example where we maximize obj_func with respect to y over 10 runs, with as many parallel processes as there are logical cores, and save progress to file.

def obj_func(x, y, pid):
    """"Function to be maximized (pid is iteration number)""""
    a = (1.5 - x + x * y)**2
    b = (2.25 - x + x * y * y)**2
    c = (2.625 - x + x * y * y * y)**2
    return -(a + b + c)
# For this example we pretend that we want to keep 'x' fixed at 0.5
# while optimizing 'y' in the range -4.5 to 4.5
pp = {'x': 0.5}  # Problem parameters
space = {'y': [-4.5, 4.5]}  # Parameters to optimize over
optimizer = GFSOptimizer(pp, space, fname="test.pkl")
# Will sample and test 'y' 10 times, then save results, progress and settings to file
optimizer.run(obj_func, n_sims=10)

For a more extensive example, see example.py.

Installation & Requirements

Requires Python >=3.6 and the following libraries:

datadiff
dlib
numpy

To install, do:

pip3 install --user gfsopt

Documentation

See example.py for an example and http://gfsopt.readthedocs.io/ for API documentation.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
gfsopt-1.0.1.tar.gz (7.6 kB) Copy SHA256 hash SHA256 Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page